This is not just a problem with formal verification systems. It's a problem intrinsic to math, science and reality.
We call these things axioms. The verifier must be assumed to be true.
For all the axioms in mathematics... we don't know if any of those axioms are actually fully true. They seem to be true because we haven't witnessed a case where they aren't true. This is similar to how we verify programs with unit tests. For the verifier it's the same thing, you treat it like an axiom you "verify" it with tests.
We call these things axioms. The verifier must be assumed to be true.
For all the axioms in mathematics... we don't know if any of those axioms are actually fully true. They seem to be true because we haven't witnessed a case where they aren't true. This is similar to how we verify programs with unit tests. For the verifier it's the same thing, you treat it like an axiom you "verify" it with tests.